[
Lists Home |
Date Index |
Thread Index
]
That's an implementation issue for searching the docs.
I have to look at Lucene. I assumed you meant loading
the schema itself. This is the tricky bit: if the
schema is not really informative about the content of
text nodes, you have to use full-text searching and
that is a very large job for any sizable collection.
deRose is the expert on this topic though.
I'm referring to a system that loads the Schema(s) in
order to create a search GUI, then let's you save
the search criteria. Ad hoc search requirements for
relational systems do this. Typical implementations
scour the table objects, get the table and field
names, then load the GUI with those. An application
specific version of that preloads those and might
attach conditions based on other rules such as security,
need to know, role-based stuff, etc. In most systems
I've worked on, one has sorted out the tables by roles
into directories, so it is easy to load Crystal by
pointing it at the directory. I've written apps
that do the same thing by querying the table objects
and creating table tables and field tables with
relationships (eg, use the table names and field names
plus their keys to do the sorting and relating). I
did it in the context of creating a schema document generator
for a relational systemm, that is, the output is a
hypertext document in multiple formats for the table
schemas to enable someone to write queries and do
conversion work. Taking those same metatables and
feeding a query generator is easy enough and adding
a topic table to that to save the queries to and feeding
a treeview object from that is a piece of cake.
Loading all of that from all of the docs is like searching
all of the available tables to get that info. Doable but
not for the faint of resources. Add full-text to that and it
becomes a job for Google farms. How well would Google
work if they weren't cacheing the web?
len
From: Robert Koberg [mailto:rob@koberg.com]
Bullard, Claude L (Len) wrote:
> Then what you want are the equivalent of
> parameterizable stored queries, somewhat
> the equivalent of the way engines like
> Crystal Reports work. Feed them a schema
> and they will give you back a UI with all of
> the queriable values from which you select,
> parameterize and store the queries. It's
> a report generator for XML documents.
But wouldn't that require loading *all* the XML docs to be searched into
memory? (I don't know Crystal Reports) (my work is in a webapp
environment). Using Lucene provides an extremely fast result set using
minimal memory.
|
